Training a Single Character style in Layer lets you teach the AI to consistently generate a specific character with the right face, expressions, outfits, poses, and personality. Whether you’re creating a main character or NPCs, this process helps lock in consistency across generations.
Start with Your Assets
For single character styles, your training images should include a variety of shots and expressions:
Headshots and closeups
 |  |
Full body poses
 |  |
Varying facial expressions
 |  |
Different angles
 |  |
This gives the model a broader understanding of your character’s visual identity: how they look, move, and emote. When you’re creating prompts later, always include the character’s name to reinforce identity.
For example:
“Dr. Pixel, smiling confidently while adjusting his goggles”
This kind of input ensures your character stays visually consistent, even as you change poses, expressions, or scenarios.
Why Captions Matter (and Why Auto-Captions Aren’t Enough)
When you upload images for style training, AI Layer will automatically caption during training, in addition to that Layer offers auto-captioning feature powered by GPT.
please review auto-captions to see if the structure of the prompt are cohesive and they only contain details about the differing aspects of the images. The less captions the better
Use an LLM to Generate a Suffix Prompt
You’ll want to kick off your LLM session with a clear prompt that defines the task, the format, the tone, and the length.
"i want you to describe the shared style among all attached images. but just the style that describes the characteristic of the style of the attached images nothing else. give me single sentence like this one : 'vector illustration with a stylized, cartoonish aesthetic, featuring bold outlines and smooth gradients'. this above one just an example dont give the same. give a new one for these uploaded images perfectly describing them. your response will be used as a prompt to generate attached image in a text to image model with various use cases so give response that way. your response will be used as a suffix.
A stylized 3D render with a glossy, polished finish, featuring vibrant colors, simple rounded shapes, and clean, bright lighting
You can enter the respective prompt in the style suffix area for AI model to incorporate this as a permanent prompt during image generation, this way you prompt AI with the art style definition it naturally sees.
You can do the same for the character prompt and enter it as a pre-fix: Prefix can be as simple as the name of the character such as Dr. Pixel
That extra attention to detail helps the model retain character identity and style much more accurately.
| Layer auto-captioning:
Chibi scientist holding a glowing, fizzing green test tube, angry expression, thick purple mustache, big goggles, bold vector art style, character design.
Manually edited captioning:
Dr. Pixel, chibi scientist holding a glowing, fizzing green test tube, angry expression, thick purple mustache, big goggles, bold vector art style, character design. |
Adding Example Prompts
Once you’ve finished your captions and uploaded your assets, you’ll reach the example prompts step, Layer auto-generates example prompts for you make sure to review them and you can always update if you have target outputs you want to see first results with
While It’s Training: Set Up Prompt Prefix + Suffix
As mentioned you can set up Prompt Prefix + Suffix during or after generation.
This helps guide how the model behaves when you generate assets, for example:
Always inserting the character’s name at the start
Always appending a style description like “3D cartoon style with soft lighting”
Prefixes and suffixes are powerful ways to lock in tone, naming, and consistency once the model is ready to use.
We’ll explore prefix/suffix best practices more in a follow-up article.